- What do you mean with "open issue"? In the academic space, the implementation and testing of constitutive models is always an issue, as the models implemented in the commercial solvers sometimes are not good enough.

- FORTRAN programming is doable but not desirable. Without experience it is horrible. Bridging the subroutine out to C/C++ is the way to go.

- There are two problems: First, the equations might be implemented incorrectly. Second, there might be programming errors. This can be very time-consuming.

- It's a good idea, but why Mathematica? Wouldn't it be more desirable to have a program which would allow you to implement the constitutive model in a smaller environment and then use the same code for an implementation into a commercial solver? What would be the target for such a program, students or more experienced FE-users?

Hi Philipp, tnx for your answer. With the open issue, I mean how much is this interesting for you and the community.

When I started in VUMAT/UMAT it first took me months to get my codes working, but only on Windows :D. Sometimes the code didn't work if I put only a new blank line. :D :D Then I started debugging things more seriously. Each model was a new story...

My idea is that for the ones who have their model, it would be much more simple to do computations in other software, quickly publish a paper and than do Fortran programming for more advanced applications.

I mentioned Mathematica because, in my opinion, it is much more simple to do scientific computations there than in other software. For example, solving nonlinear equations, performing matrix manipulations, integration, convolution and many other operations can be done in a short time without special effort. Using "mathematical" software should make things easier, I guess...

What do you think?

Hi,

i have obtained pretty good experience with open source Fenics + MFront technologies to quick testing constitutive laws and models formulated with PDEs. It is written in Python, which is high level easy language. The Fenics philosophy is to write equations in a form as close as much to a mathematical one!

If you are only interested in testing the constitutive model using simple numerical examples, then it would be best be to write your own FEM code.

The programming language of choice is entirely personal; the best language is the one you are familiar with. Matlab or Python, even Mathematica, is good enough for this case.

However, things become complicated if you want to apply those constitutive models for complex large-scale problems, for example, contacts, fracture, etc. For such cases, an established codebase written in C/C++/Fortran would be a better choice. Python & Matlab are significantly slower compared to C/C++/Fortran.

I use Matlab for prototyping tensorial derivatives of constitutive models and C++ for FEA simulations.

I am happy to discuss further if you are interested.